Research On Automatic Optical Inspection Method For Defects In Array Manufacturing Process Of Display Panel

By 2019,the mainland of China will account for nearly 40% of global display panel production capacity.However,the Array segment process equipment,which accounts for 75% of the total equipment investment,is almost monopolized by Japan,South Korea and the United States.Optical Automatic Inspection(AOI)equipment is one of the core equipment of the Array segment process of large-size display panels,which directly determines the yield of the product.Its main purpose is to detect all kinds of pixel defects,mask defects,line defects and foreign body defects in the manufacturing process,and provide information on the exact location,size and type of defects,so as to improve the manufacturing process and guide defect repair,thereby improving product yield.The multi-layer film structure of the panel and the complex process of the Array manufacture process cause the texture of the substrate to be complex,and there are a large number of near-transparent,low-contrast micro-defects,which brings great difficulties and challenges to defect detection.In response to the above problems,this subject carried out "Research on Automatic Optical Inspection Method for Defects in Array Manufacturing Process of Display Panel".The research works are as follows:(1)Carried out the research on the uneven illumination correction algorithm based on polynomial curve fitting and Retinex theory.On the basis of completing the design of the micro-defect optical automatic detection system,the effect of uneven illumination of the system imaging on the defect detection was discussed.Based on the curve fitting and multi-scale Gaussian function,carried out a comparative study of the uneven illumination correction algorithm.The correction method using the combination of quadratic and quartic polynomials curve fitting improves the uneven illumination correction result;according to the light component estimation model based on Retinex theory,the light components extracted at different scales using the multi-scale Gaussian function are used for eliminating the uneven illumination;establising the gray uniformity evaluation metric function,combined with the effect of the correction algorithm on local defects,and make a comprehensive comparison evaluation of the uneven illumination correction results.Based on the evaluation result,the multi-scale Gaussian function correction algorithm can achieve a better correction of theimage to be inspected in this subject.(2)Carried out a research on periodic self-difference defect detection algorithm.For the problem of difficult detection of micro-defects in the background of periodic complex textures,on the basis of the 5-point neighborhood sorting comparison method and the periodic similarity of the image,a periodic self-differential shadow defect detection algorithm is proposed,which overcomes the defect that the sorting comparison method is too sensitive and the image edge cannot be detected.The proposed method can achieve a defect detection rate of93.2% on the test sample group,and there is no false detection in the detected defects.And a defect rating evaluation method is given,which provides a priority criterion for the defect re-examination.(3)Carried out research on low-contrast micro-defect detection methods based on sub-image error templates and image enhancement.Introduced sub-image error templates in periodic self-difference defect detection algorithm to reduce false detection points.The combination of the nonlinear transformation of low gray value stretch and the sub-image error template solves the problem of low-contrast defect detection,and 100% defect detection can be achieved on the test sample group.On this basis,one-dimensional and two-dimensional distance matching functions are used to realize the measurement of image regularity,and then automatically determine the lengths of the image periods.The functions reduce manual parameter input and improve the degree of automatic detection by the algorithm.